Oceanologia (2016) 58, 90—102
Available online at www.sciencedirect.com
ScienceDirect
jou rnal homepage: www.elsevier.com/locate/oceano
ORIGINAL RESEARCH ARTICLE
Curonian Lagoon drainage basin modelling and
§
assessment of climate change impact
a b,a a c
Natalja Čerkasova , Ali Ertürk , Petras Zemlys , Vitalij Denisov , d,a,
Georg Umgiesser *
a
Open Access Centre for Marine Research, Klaipeda, Lithuania
b
Department of Freshwater Biology, Istanbul University, Istanbul, Turkey
c
Faculty of Marine Technology and Natural Sciences, Klaipeda University, Klaipeda, Lithuania
d
ISMAR-CNR, Institute of Marine Sciences, Venezia, Italy
Received 8 September 2015; accepted 12 January 2016
Available online 9 February 2016
KEYWORDS Summary The Curonian Lagoon, which is the largest European coastal lagoon with a surface
2 2
area of 1578 km and a drainage area of 100,458 km , is facing a severe eutrophication problem.
Drainage basin
modelling; With its increasing water management difficulties, the need for a sophisticated hydrological
SWAT; model of the Curonian Lagoon's drainage area arose, in order to assess possible changes resulting
from local and global processes. In this study, we developed and calibrated a sophisticated
Curonian Lagoon;
hydrological model with the required accuracy, as an initial step for the future development of a
Nemunas basin;
modelling framework that aims to correctly predict the movement of pesticides, sediments or
Climate change
nutrients, and to evaluate water-management practices. The Soil and Water Assessment Tool was
used to implement a model of the study area and to assess the impact of climate-change scenarios
on the run-off of the Nemunas River and the Minija River, which are located in the Curonian
Lagoons drainage basin. The models calibration and validation were performed using monthly
2
streamflow data, and evaluated using the coefficient of determination (R ) and the Nash-Sutcliffe
2
model efficiency coefficient (NSE). The calculated values of the R and NSE for the Nemunas and
Minija Rivers stations were 0.81 and 0.79 for the calibration, and 0.679 and 0.602 for the
validation period. Tw o potential climate-change scenarios were developed within the general
§
This study was funded by the European Social Fund under the Global Grant measure (CISOCUR project VP1-3.1-ŠMM-07-K-02-086).
* Corresponding author at: ISMAR-CNR, Institute of Marine Sciences, Arsenale — Tesa 104, Castello 2737/F, 30122 Venezia, Italy.
Tel.: +39 339 4238653.
E-mail address: [email protected] (G. Umgiesser).
Peer review under the responsibility of Institute of Oceanology of the Polish Academy of Sciences.
http://dx.doi.org/10.1016/j.oceano.2016.01.003
0078-3234/# 2016 Institute of Oceanology of the Polish Academy of Sciences. Production and hosting by Elsevier Sp. z o.o. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Curonian Lagoon drainage basin modelling and assessment of climate change impact 91
patterns of near-term climate projections, as defined by the Intergovernmental Panel on Climate
Change Fifth Assessment Report: both pessimistic (substantial changes in precipitation and temper-
ature) and optimistic (insubstantial changes in precipitation and temperature). Both simulations
produce similar general patterns in river-discharge change: a strong increase (up to 22%) in the
winter months, especially in February, a decrease during the spring (up to 10%) and summer (up to
18%), and a slight increase during the autumn (up to 10%).
# 2016 Institute of Oceanology of the Polish Academy of Sciences. Production and hosting by Elsevier
Sp. z o.o. This is an open access article under the CC BY-NC-ND license (http://creativecommons.
org/licenses/by-nc-nd/4.0/).
2010). As a result, the lagoon's water level is usually higher
1. Introduction
than that of the Baltic Sea; therefore, the dominant currents
0 are from the lagoon to the Baltic Sea.
The Curonian Lagoon is located at N 55830 latitude and
0 Over the years, the water discharge to the lagoon chan-
E 21815 longitude. It is the largest European coastal lagoon,
ged, and this led to a fluctuation of the water balance. Major
separated from the Baltic Sea by a narrow 0.5—4 km wide
changes have been observed in the last decade in the winter—
sandy Curonian spit and connected to the Baltic Sea through
spring period. In the winter months of January and February,
the Klaipeda Strait. Several small rivers — such as the Bol-
due to observed warmer winters, the Nemunas' run-off has
shaya and Malaya Morianka, Kalinovka, Deima, Rybnaya,
increased, while spring floods are decreasing; therefore, run-
Minija, Dane and Dreverna — and one large river (Nemunas)
off levels over the year became more homogeneous (Žilinskas
discharge into the Curonian Lagoon. The southern and central
et al., 2012).
parts of the lagoon contain fresh water due to the discharge
Agriculture has a significant impact on the status of water
from those rivers. The run-off of rivers to the lagoon varies
3 3 1 3 1 bodies in the Nemunas River basin, especially in the sub-basins
from 14 to 33 km per year (444 m s to 1046 m s ) and
of the Sesupe and Nevezis Rivers; this factor has a local, but
exhibits a strong seasonal pattern, peaking with snowmelt
serious, impact. Chemicals that enter the river from agricul-
during the flood season of March to April (Dubra and
ture and fishponds are a major source of pollution. A substan-
Červinskas, 1968).
tial proportion of point-source pollution comes from industry.
The area of land draining into the Curonian Lagoon covers
2 According to the Second Assessment of Transboundary Rivers,
100,458 km , of which 48% lies in Belarus, 46% in Lithuania,
Lakes and Groundwaters by the United Nations Economic
and 6% in the Kaliningrad Oblast and Poland (Gailiušis et al.,
Commission for Europe (2011), there is room for development
1992) (see Fig. 1). The drainage area of the Curonian Lagoon
in the monitoring of the Nemunas River, as the current list of
consists of several river basins; however, the most important
monitored pollutants is limited. There is a lack of biological
of them is the Nemunas River drainage basin in terms of flow
observation and monitoring of pollutants in river-bottom sedi-
rates and nutrient inputs, supplying about 98% of its inflows
ments, and a joint, harmonized monitoring programme for the
(Jakimavičius, 2012). The annual Nemunas River water
transboundary watercourses is needed. It is important to
inflow into the Curonian Lagoon is more than three times
develop a model for nutrient and other biogeochemically
greater than the volume of water in the lagoon (Žilinskas
significant dissolved-substance contributions that are altering
et al., 2012). According to researchers, the average annual
3 3 1 and influencing the ecosystems of the Nemunas River and
run-off during 1812 to 2002 was 22.054 km (699 m s )
Curonian Lagoon.
(Gailiušis et al., 1992), and from 1960 to 2007 it was
3 3 1 The first step is the development of a hydrological model
21.847 km (692 m s ) (Jakimavičius and Kovalenkovienė,
and analysis of changes in the Curonian Lagoon drainage basin
due to global processes (climate change, etc.), as well as
local anthropogenic activities, and forecasting possible
changes in the future. Model hydrology calibration, uncer-
tainty analysis and sensitivity analysis enables a broader
understanding of key processes in the catchment area.
Recent work conducted in the field of Curonian Lagoon
drainage basin modelling is reported in the doctoral disserta-
tion “Changes of water balance elements of the Curonian
Lagoon and their forecast due to anthropogenic and natural
factors” by Jakimavičius (2012). In this work, the author had
created hydrological models for the separate Nemunas catch-
ment areas using HBV (Hydrologiska Byråns Vattenbalansav-
delning), before calibrating and validating them. The
sensitivity and uncertainty of the Nemunas run-off model
parameters were assessed using the SUSA (Software for
Uncertainty and Sensitivity Analyses) package. Hydrometeor-
ological information of the period 1961—1990 was used for
the model's creation. The period 1961—1975 was selected
for the model's calibration, whereas the period 1976—1990
Figure 1 Curonian Lagoon drainage area.
92 N. Čerkasova et al.
was used for its validation. Prognostic data from 14 measure- model that operates on a daily time step and is designed
ment stations, data from 1961 to 1990 and the downscaling to predict the impact of management on water, sediment,
method were applied to calculate the daily mean data from and agricultural chemical yields in ungauged drainage areas
the mean monthly output data of climate-change scenarios. (Arnold et al., 1993).
In this way, obtained prognostic values of precipitation and The drainage basin area was divided into multiple sub-
temperature data were used to simulate the Nemunas' inflow basins, which were then further subdivided into Hydrological
and to compute its water balance (Jakimavičius, 2012). Response Units that consist of homogeneous land-use, man-
Projections of the temperature and precipitation of the agement and soil characteristics (Arnold et al., 1993). Using
Nemunas' river basin for the 21st century (according to the SWAT, run-off was predicted for each HRU separately and
conclusions of the Fourth Assessment Report of the United routed to obtain the total run-off of a catchment area. This
Nations Intergovernmental Panel on Climate Change, as well solution improves the model's accuracy and provides a much
as the results of output data of ECHAM5 and HadCM3 global better physical description of the water balance (Neitsch
climate models under A2, A1B and B1 greenhouse gas emis- et al., 2011).
sion scenarios) were used to create the climate-change A stable version of the SWAT from 2009 was used due to the
scenarios. These data were used to compute the Nemunas' fact that additional extensions and extra tools are available
inflow to the lagoon during 2011—2100, the amount of pre- for this version, and also because this version has undergone
cipitation entering the lagoon and water evaporation from much testing and correction.
the lagoon's surface (Jakimavičius, 2012). For the calibration of the Curonian Lagoon drainage basin
The study conducted by Jakimavičius (2012) is quite model, the SWAT-CUP (Soil and Water Assessment Tool Cali-
comprehensive; however, it lacks some key points. The bration and Uncertainty Programs) semi-automated SUFI-2
research mainly focuses on the changes of water-balance method (Sequential Uncertainty Fitting, version 2) was
elements of the Curonian Lagoon, such as the Nemunas applied, as it is most commonly used, well-documented
River's inflow to the lagoon, and the water exchange between method, and is reported to produce satisfactory results
the Baltic Sea and the Curonian Lagoon, with no focus on the (Arnold et al., 2012). Model calibration and validation were
smaller rivers', such as the Minija, run-off change. The pre- performed using monthly streamflow data.
cipitation amount used covered only the territory of Lithua- Several publications (Arnold et al., 2012; Balascio et al.,
nia. The selected baseline period is outdated and does not 1998; Moriasi et al., 2007) have examined the usage of
fully represent the current conditions of the catchment area. different model-evaluation statistics; however, not many
The climate change scenarios covered the precipitation of them provide directions or advice on using acceptable
amount and temperature change, with no change to the ranges of values for these performance indicators. Some of
relative humidity. the guidelines suggest using the NSE and the coefficient of
2
The SWAT (Soil and Water Assessment Tool) model is also determination (R ), in addition to graphical techniques (Mor-
used by Lithuania's Ministry of Environment in development iasi et al., 2007). Suggested guidelines were followed, and
of a method and modelling system for nitrogen and phos- the Curonian Lagoon's drainage basin model was evaluated
phorus load-calculation for the surface waters of Lithuania according to them.
(ELLE and PAIC, 2012). The model covers only the territory of
Lithuania, which is divided into more than 1200 sub-basins.
2.1. SWAT model description and features
The developed model uses high-resolution DEM (Digital
Elevation Model), soil and land-use data layers in order to The development of the SWAT is a continuation of the United
create Hydrologic Response Units (HRUs) with a resolution of States Department of Agriculture Agricultural Research Ser-
5 m 5 m; therefore, the model's accuracy and predictive vice (USDA-ARS), a modelling experience that spans a period
capability is reduced. Overall, the model's Nash-Sutcliffe of roughly 30 years. The current SWAT model is a direct
efficiency coefficient (NSE) performance for the monthly descendant of the Simulator for Water Resources in Rural
median flow is 0.5. The model is primarily used for the Basins (SWRRB) model, which was designed to simulate
development of methods and tools for multi-objective spatial management impacts on water and sediment movement
optimization, and structural agriculture-change scenario for ungauged rural basins across the U.S. (Arnold et al.,
assessments. With additional model set-up corrections and 2010). The SWAT has experienced constant reviews and an
a more thorough calibration, this model can become state-of- extension of its functionality since it was created in the early
the-art for Lithuania's territory; however, it is not open 1990s. The most significant improvements are listed in the
access, is unavailable for usage outside of the Lithuanian official SWAT theoretical documentation (Neitsch et al.,
Environmental Protection Agency and results have not yet 2011) and include the following: incorporation of multiple
been published in peer-reviewed journals. These facts led to HRUs, auto-fertilization and auto-irrigation management
the necessity of producing a more flexible tool for analysing options, incorporation of the canopy storage of water, the
and predicting hydrological and biogeochemical cycles of the Penman-Monteith potential evapotranspiration equation, in-
Curonian Lagoon's drainage basin. In addition, the created stream water quality equations, improvement of snow melt
model could allow the exchange of modelling results and routines, nutrient cycling routines, rice and wetland rou-
benefit development of large-scale modelling systems. tines, bacteria transport routines, Green and Ampt infiltra-
tion, weather generator improvements and many other
2. Material and methods factors. The incorporation of the Curve Number (CN) method
and non-spatial HRUs allow adaptation of the model to
The Curonian Lagoon drainage basin model was set up using virtually any drainage basin with a wide variety of hydro-
the SWAT: a physically based, continuous-time catchment logical conditions (Gassman et al., 2007). Simulation of the
Curonian Lagoon drainage basin modelling and assessment of climate change impact 93
hydrology of a catchment area can be separated into two Spatial Information, accessed: February 2014) based on
major points: the SRTM (Space Radar Topographic Mission) survey data
provided by NASA (National Aeronautics and Space Adminis-
1. Land phase of the hydrological cycle: the amount of tration) dating back to 2001. The resolution of the DEM is
water, nutrient, sediment and pesticide loadings in the 51 m 51 m. To use this DEM in the SWAT, the grid size had to
main channel in each sub-basin. be changed to a coarser resolution. This decision is based on
2. Water or routing phase of the hydrological cycle: the studies of the DEM's resolution effects on the SWAT's output.
movement of water, sediments, etc., through the chan- Studies showed that run-off had little or no sensitivity to the
nel network of the drainage area to the outlet (Neitsch resampled resolution, and the minimum DEM resolution
et al., 2011). should range from 100 m to 200 m in order to achieve a less
than 10% error rate in the SWAT's output for flow, NO3-N and
The hydrologic cycle simulated by SWAT is based on the water total P predictions (Chaubey et al., 2005; Ghaffari, 2011; Lin
balance equation: et al., 2010). A coarser resolution results in a decrease of the
computational needs by up to three times during model's set-
Xt
up and run phases. The DEM grids were resampled to a size
¼ þ ð Þ;
SW tþ1t SWt0 Rday Q surf Ea wseep Q gw (1)
i¼1 of 153 m x 153 m, which is within the recommended range
(Chaubey et al., 2005).
where SWt+1t is the final soil water content at day t [mm H2O],
SWt0 the initial soil water content, Rday the amount of
precipitation on day t [mm H2O], Qsurf the amount of surface 2.2.2. Land-use data
runoff on day t [mm H2O], Ea the amount of evapotranspira- Land-use data were acquired from the WaterBase project
tion on day t [mm H2O], wseep the amount of water entering database (United Nations University, accessed: November
vadose zone from the soil profile on day t [mm H2O], and Qgw 2014) based on the FAO's (Food and Agriculture Organization
is the amount of return flow on day t [mm H2O]. of the United Nations) land-use data. There are 13 classes of
land-use types in the study area, which correspond to the
ones used in the SWAT database (Table 1). The most dominant
2.2. Model set-up and data
type of land-use in the area is CRWO (covering 64% of the
study area), which is the abbreviation for “cropland, wood-
There are four main data sets that were used in the SWATset-
” — “
up: land mosaic , followed by CRDY (23%) dryland, cropland
and pasture” and FOMI (6%) — “mixed forest”. The informa-
tion required to simulate plant growth is stored in the SWAT
1. Digital Elevation Model (DEM) data;
plant-growth database file according to plant species. The
2. Land-use data;
SWAT uses a plant-growing cycle in order to determine how
3. Soil data;
much water is consumed by the canopy, and how much can be
4. Weather data.
stored and released by it. The model takes into account
growing seasons, harvesting and other parameters, which
There are many more optional datasets that can be used as
can be specified or modified by the user during the model's
inputs for the SWAT.
set-up stage.
2.2.1. DEM
2.2.3. Soil data
DEM data were obtained from the Consortium for Spatial
Soil data were acquired from the WaterBase project database
Information (CGIAR-CSI) database (CGIAR — Consortium for
of United Nations University (accessed: November
2014). Twenty-six classes of soil are present in the study
Table 1 Land use type occurrence in the Curonian Lagoon
area, which correspond to the ones used in the SWAT data-
drainage basin area.
base (Table 2). Soil-class characteristics can be determined in
the same way as for land-use classes, by using a database that
Nr. Class Land use type Area
is implemented in the tool, which contains the most common
label [% of total]
soil types and their properties.
1 CRWO Cropland/woodland mosaic 64 Soil data used by the SWAT can be divided into two
2 CRDY Dryland cropland and pasture 23 groups: physical characteristics and chemical characteris-
3 FOMI Mixed forest 6 tics. Physical properties of the soil govern the movement of
4 CRGR Cropland/grassland mosaic 3 water and air through the soil's profile, and have a major
5 WATB Water bodies 2 impact on the cycling of water within the HRU, whereas
6 FOEN Evergreen needleleaf forest 2 inputs for chemical characteristics are used to set the
7 FODB Deciduous broadleaf forest 1 initial levels of chemicals that are present in the soil
<
8 URMD Residential medium density 1 (Neitsch et al., 2011). Physical properties for each soil
<
9 GRAS Grassland 1 type are necessary for use of the model, while chemical
<
10 FODN Deciduous needleleaf forest 1 ones are optional. The most widely presented soil layer in
<
11 CRIR Irrigated cropland and pasture 1 the study area is De18-2a-3049 (33%), which is categorized
<
12 TUWO Wooded tundra 1 as “loam”, followed by Gm32-2-3a-3074 (10%) and Lg55-1a-
<
13 SHRB Shrubland 1 31993199 (9%), which are “clay loam” and “sandy loam” respectively.
94 N. Čerkasova et al.
by the Klaipeda University Coastal Research and Planning
Table 2 Soil class occurrence in the Curonian Lagoon drain-
Institute (KU CORPI).
age basin area.
Nr. Class label Soil texture Area
2.2.6. Sub-basin set-up
[% of total]
The study area was delineated using the MapWindow Terrain
Analysis Using Digital Elevation Models (TauDEM) tool, follow-
1 De18-2a-3049 LOAM 33
ing the specification of the threshold drainage area, which
2 Gm32-2-3a-3074 CLAY_LOAM 10
is the minimum drainage area required to form the origin
3 Lg55-1a-3199 SANDY_LOAM 9
of the stream, and identification of the drainage basin outlets.
4 De20-2ab-3052 LOAM 8
The accuracy of the delineation process was influenced by the
5 De18-1a-3048 SANDY_LOAM 6
DEM's resolution. Several iterations were performed with
6 De17-1-2a-3047 SANDY_LOAM 4
different delineation threshold-values, thus creating models
7 Pl5-1ab-3236 LOAMY_SAND 4
with different numbers of sub-basins. Each model's initial
8 Lo78-1-2a-3204 SANDY_LOAM 3
output performance was tested in order to analyse the deli-
9 Be144-2-3-3019 CLAY_LOAM 3
neation threshold-value's influence on its predictive flow cap-
10 Dg5-1ab-3055 SANDY_LOAM 3
abilities. As a result, a total of 117 sub-basins were produced,
11 Dd8-1ab-3045 SANDY_LOAM 2
which proved to be the best performing number of sub-basins
12 De13-1ab-3046 SANDY_LOAM 2
for this study. Multiple HRUs were then created automatically
13 De19-1a-3050 SANDY_LOAM 2
with the MapWindow SWAT plug-in within each sub-basin, as a
14 De19-2a-3051 SANDY_LOAM 1
function of the dominant land-use and soil types.
15 Je87-2-3a-3149 CLAY_LOAM 1
16 Lg41-2-3a-3194 LOAM 1
17 Lg43-2ab-3196 SANDY_LOAM 1 2.3. Calibration procedure
18 Lo69-2ab-3201 LOAM 1
—
19 Od22-a-3217 LOAM 1 The available period of observation data (2000 2010) was
—
20 Oe14-a-3223 LOAM 1 divided into the two groups of 2000 2007 for calibration and
—
21 Pl5-1ab-3236 LOAMY_SAND <1 2008 2010 for validation; this supplies a period of 8 years for
22 Po30-1ab-3239 SANDY_LOAM <1 calibration and 3 years for validation.
23 Be126-2-3-6436 LOAM <1 Various studies have reported different input parameters
24 Lo79-2a-6572 LOAM <1 used in the SWAT model's calibration. Table 3 summarizes the
25 Lo81-1a-6574 SANDY_LOAM <1 most frequently used parameters in various studies (Abbas-
26 Qc62-1a-6623 SAND <1 pour, 2011; Arnold et al., 2012). As the SWAT is a comprehen-
sive model that simulates process interactions, many
parameters will impact multiple processes. For instance, CN
(Curve Number) directly impacts surface run-off; however, as
2.2.4. Weather data
surface run-off changes, all components of the hydrological
Historical weather data are usually gathered and archived by
balance change. The described feature is the primary reason
the countries' meteorological services. In Lithuania, such
for calibrating the model starting with the hydrological bal-
data are available from the Lithuanian Hydrometeorological
ance and streamflow, then moving to sediment and, finally,
Service, under the Ministry of Environment of the Republic of
calibrating nutrients and pesticides (Arnold et al., 2012).
Lithuania (LHMS). However, the study area covers not only
All suggested parameters described in Table 3 were subjected
the territory of Lithuania, but also Kaliningrad Oblast and
to calibration and sensitivity analysis in order to regionalize
Belarus, meaning that data had to be acquired from global
the most sensitive parameters and make the necessary adjust-
public resources. The weather data were acquired through
ments to their values. These steps were performed iteratively,
Global Weather Data for the SWAT service (National Centers
as recommended in the SUFI-2 calibration procedure docu-
for Environmental Prediction, Accessed: November 2014). It
mentation (Abbaspour, 2011). The maximization of NSE for
provides data for the 35-year period between 1979 and
river discharge was used as an objective function:
2014. The service allows the downloading of daily data for P
2
precipitation, wind, relative humidity and solar radiation in ð Þ
i Q m Q s i
NSE ¼ 1 P ; (2)
fi the SWAT's le format for a given location and time period. ð Þ2
i Q m;i Q m
Weather data used for the Curonian Lagoon drainage basin
model covered a period of 16 years, from 1995 to 2010; the where Qm is the measured parameter value (e.g. river dis-
fi fi —
rst ve years' data (1995 1999) were used for the model's charge), Qs the simulated parameter value, and Q m is the
warm-up stage, whereas the remaining data were used for average value of measured parameter. NSE values ranges
the model's set-up, calibration and validation. between 1 and 1.0, with 1.0 being the optimal value. Values
between 0.0 and 1.0 are generally viewed as acceptable levels
2.2.5. Observed data of performance, whereas values below 0.0 indicate unaccept-
The observed run-off data files had to be prepared for the able performance of the model (in this case the mean observed
model's output analysis and calibration. The available value is a better predictor than the simulated one) (Krause
observed data were for two river discharges: Nemunas, near et al., 2005).
Smalininkai, and Minija, near Lankupiai (Fig. 2). The data SWAT-CUP calibration iteration presumes the existence
files present the daily time series over 11 years (2000—2010) of a set of simulations with predefined parameters, uncer-
3 1
for measured discharges in m s . These data were provided tainty ranges for these parameter values, statistics of every
Curonian Lagoon drainage basin modelling and assessment of climate change impact 95
Figure 2 Minija and Nemunas river historical discharge.
simulation output and overall statistics of the calibration Some parameter values are similar for every sub-basin,
process. A simulation is a single execution of the SWAT model, whereas others differ substantially. This can be explained
with certain parameter values within a boundary of the by the spatial distribution of sub-basins and their differ-
uncertainty ranges, as defined in the iteration set-up pro- ences in soil, land-use and the topographic properties of
cess. The initial calibration run was carried out with the area. Since sub-basins and HRUs are spatial averaging
2000 model simulations. As the number of parameters and over some area, the parameter values for the same catch-
simulations was high, the time consumption of such a cali- ment area will change as the sizes of sub-basins and HRUs
bration iteration was demanding. To complete one iteration change.
cycle on a standard laptop computer (with an Intel Core i7 Defining proper parameter boundaries for parameters
2.4 GHz processor) with 2000 simulations, the calibration used in the calibration stage can be a challenging process.
tool had to run for 36 h (for a model with 117 sub-basins). These ranges have a strong impact on the autocalibration
The speed performance of the calibration tool is strongly outcome. In some SWAT model autocalibration studies
dependent on the complexity of the model (the number of (Arnold et al., 2012; Balascio et al., 1998; Moriasi et al.,
sub-basins, HRUs, calibration parameters and simulated per- 2007), different parameter ranges are used for the same
iod), the processor's clock speed and some other factors, such parameters that are subjected to calibration, but the expla-
as the speed of the hard drive and the architecture of the nation for such boundary usage is not always provided.
machine. Studies have shown that the calibration procedure's The high number of parameters complicates and prolongs
run time could be enhanced by enabling the parallel proces- the model's parameterization and calibration procedure, and
sing of the calibration tool (Rouholahnejad et al., 2012). can therefore be considered as a weakness of the model,
Calibration was carried out, accounting for spatial para- especially if the soil and geological differences of the catch-
meter variations in different basins (of the Minija and Nemunas ment area are not well known. This was the drawback for
Rivers) (see Fig. 3). Global model performance for the monthly Curonian Lagoon basin model, as the soil and land-use data
2
run-off values of NSE = 0.79 and R = 0.81 were achieved, were acquired from public sources and not from local ones,
which correspond to very good ratings (see Table 4). This which would be of better quality and backed-up by more
model was subjected to further validation and used in the recent observations. Different competences in various fields
scenario assessment. SWAT-CUP produces a fitted parameter of study are required in order to fully assess the influence of
value table for the best simulation. Corresponding values are each parameter and its value to the basin. A more detailed
given in Table 5. analysis of each parameter, not only those that are used in
96 N. Čerkasova et al.
Table 3 Most frequently used calibration parameters in various SWAT model calibration studies.
Parameter Definition Process
CN2 Initial Soil Conservation Service runoff curve number Surface runoff
1
CH_K1 Effective hydraulic conductivity in tributary channel alluvium [mm h ]
1
CH_K2 Effective hydraulic conductivity in main channel alluvium [mm h ]
CH_N2 Manning's “n” value for the main channel
ESCO Soil evaporation compensation factor
EPCO Plant uptake compensation factor
SURLAG Surface runoff lag coefficient
CANMX Maximum canopy storage [mm H2O]
ALPHA_BF Baseflow alpha factor [days] Baseflow
GW_REVAP Groundwater “revap” coefficient
GW_DELAY Groundwater delay time [days]
GWQMN Threshold depth of water in the shallow aquifer required for return flow to occur [mm H2O]
GWHT Initial groundwater height [m]
REVAPMN Threshold depth of water in the shallow aquifer for “revap” or percolation
to the deep aquifer to occur [mm H2O]
SFTMP Snowfall temperature [8C] Snow
1
SMFMN Melt factor for snow on December 21 [mm H2O 8C-day ]
1
SMFMX Melt factor for snow on June 21 [mm H2O 8C-day ]
SMTMP Snowmelt base temperature [8C]
TLAPS Temperature laps rate [8C]
SOL_Z Depth from soil surface to bottom layer [mm] Soil
SOL_AWC Available water capacity of the soil layer
SOL_ZMX Maximum rooting depth of soil profile [mm]
3 3
SOL_BD Moist bulk density [Mg m ] or [g cm ]
1
SOL_K Saturated hydraulic conductivity [mm h ]
SOL_ALB Moist soil albedo for top layer
Figure 3 Calibration results for Minija and Nemunas river discharge.
Curonian Lagoon drainage basin modelling and assessment of climate change impact 97
Table 4 Statistics report for the calibration period (2000—2007) of the Curonian Lagoon drainage basin model.
2 3 1 3 1
Variable P-factor R NSE Mean (simulated) [m s ] StdDev (simulated) [m s ]
FLOW_OUT (Minija) 0.28 0.77 0.76 32.61 (30.59) 29.99 (23.91)
FLOW_OUT (Nemunas) 0.38 0.85 0.84 455.91 (470.84) 204.85 (208.56)
this study, would benefit future developments of a Curonian The global Curonian Lagoon drainage area model validation
2
Lagoon drainage-basin model. results are R = 0.679 and NSE = 0.602, which correspond to
satisfactory values for the model at a monthly time-step. The
validation result confirms that this area-specific hydrological
2.4. Validation procedure
model can produce sufficiently accurate predictions.
Although the required model performance objectives
Validation results show whether the parameters were cali-
were met, validation results give an insight into possible
brated in such a way as to represent the modelled system
errors in the output. The SWAT model was unable to predict
adequately, in this case, the Curonian Lagoon's drainage
the high amounts of run-off occurring in the spring months
area. The model's validation output can be analysed in the
2 (for both observation points), and some peak flows were
same way as the model calibration: a value for R and NSE can
underestimated. The model failed to simulate the high
be computed and the plot of the simulated flow, as compared
amount of run-off that occurred in the late autumn and
to the observed flow, produced. After the successful valida-
winter months (November—January) of 2008—2009 and
tion of the model, it can be used for various purposes:
2009—2010 for both Minija and Nemunas (see Fig. 4). For
monitoring seasonal and long-term trends, predicting any
improving the model's predictive accuracy, snowmelt and
of the model's output elements under different conditions
ice-formation parameter temperatures might be adjusted
and scenarios and using outputs as inputs for other models.
to account for early melting or late ice/snow formation.
Table 5 Fitted parameter values for the Curonian Lagoon
drainage basin model. 2.5. Scenario formulation
a
Parameter Fitted parameter values
Air temperature and the precipitation amount are the main
Nemunas sub-basin Minija sub-basin climate elements directly affecting the total run-off of
b c rivers. Prognostic air temperature, precipitation amount
R _CANMX.hru 24.810646 39.028244
and humidity-change data, derived from the Intergovern-
R_CH_N2.rte 27.378458 37.603989
mental Panel on Climate Change Fifth Assessment Report
R_CN2.mgt 248.030533 43.943916
(IPCC AR5), were used with the SWAT for modelling river
R_SOL_ALB.sol 0.964783 19.724804
inflow changes.
R_SOL_AWC.sol 44.375572 71.344315
Near-term projections from the General Circulation
R_SOL_BD.sol 0.63497 1.009578
Model-Regional Climate Model (GCM-RCM) model chains
R_SOL_K.sol 10.870995 0.729487
for Europe were used for modelling precipitation and tem-
R_SOL_Z.sol 1.501782 30.346178
perature changes. The analysis includes the following
R_SOL_ZMX.sol 36.707863 35.845459
10 GCM-RCM simulation chains for the Special Report on
V_ALPHA_BF.gw 0.116285 1.178193
Emissions Scenarios' (SRES) A1B scenario (which includes
V_CH_K1.sub 199.661301 63.989532
the RCM group and GCM simulation): HadRM3Q0-HadCM3Q0,
V_CH_K2.rte 11.417052 53.535545
ETHZ-HadCM3Q0, HadRM3Q3-HadCM3Q3, SMHI-HadCM3Q3,
V_EPCO.bsn 0.651629 0.598089
HadRM3Q16- HadCM3Q16, SMHIBCM, DMI-ARPEGE, KNMI-
V_ESCO.bsn 0.484453 0.799542
ECHAM5, MPI-ECHAM5, DMI-ECHAM5 (Kirtman et al., 2013).
V_GW_DELAY.gw 279.003998 72.274193
The CMIP5 (Coupled Model Intercomparison Project, Phase 5)
V_GW_REVAP.gw 0.143441 0.11155
multi-model ensemble was used for the relative humidity
V_GWHT.gw 14.892682 4.632501
change.
V_GWQMN.gw 381.135162 803.918335
The current-condition scenario was carried out before
V_REVAPMN.gw 331.778961 18.785471
implementation of the climate-change scenario simulations;
V_SMFMN.bsn 3.019845 9.110275
the produced average monthly run-off values were consid-
V_SMFMX.bsn 12.965604 9.025815
ered as the baseline. In order to analyse the impacts of
V_SMTMP.bsn 0.340758 0.831538
potential future climate change on the hydrology of the
V_SURLAG.bsn 27.429886 8.502431
Curonian Lagoon drainage area, every scenario was imple-
V_TLAPS.sub 2.980006 8.614986
mented with downscaled, spatially variable climate inputs
a
Parameter definitions are given in Table 3. (air temperature precipitation, relative humidity) using the
b
fi
The quali er (V_) refers to the substitution of a parameter by a matching simulation period, which delivers a consistent
value from the given range, while (R_) refers to a relative change foundation for comparison of the scenario outputs. The
in the parameter where the current value is multiplied by 1 plus a
near-term change and projected changes described are for
factor in the given range.
—
c the period 2016 2035.
The extension (.hru,.rte,.mgt,.sol,.gw,.sub,.bsn) refer to the
According to the summary of IPCC AR5, air temperature is
SWAT file type where the parameter occurs.
going to increase by up to 18C in winter, 0.58C in spring and
98 N. Čerkasova et al.
Figure 4 Validation result with the best fitted parameter value set for Minija and Nemunas river discharge.
autumn and 0.78C in summer in the near-term; precipitation conditions of climate change, and in order to determine the
is expected to increase by 7.5% in winter, 5.0% in autumn and sensitivity of the modelled system.
spring, and 2.5% in summer and humidity is likely to decrease
slightly, by about 1% over most land areas (Kirtman et al.,
3. Results
2013). Tw o climate change scenarios, one pessimistic and one
optimistic, were formulated (Table 6), and the effects of
3.1. The pessimistic scenario's results
these scenarios on river run-off were explored. The pessi-
mistic scenario includes high values for temperature and For the pessimistic scenario, high values of precipitation and
precipitation change, whereas the optimistic scenario's cor- temperature change were used to assess their effects on the
responding values were lower. Such scenarios were formu- Nemunas and Minija Rivers' run-off. For the Nemunas River,
lated for assessing the response of the study area to various the projected changes in precipitation, temperature and
Table 6 Climate variable change in pessimistic and optimistic scenarios.
Scenario Simulated changes in:
Temperature [8C] Precipitation [%] Relative humidity [%]
a b c
DJF MAM, SON JJA DJF MAM, SON JJA All seasons
Pessimistic +2 +0.5 +0.7 7.5 5 2.5 1
Optimistic +0.6 +0.4 +0.3 3 1.5 0 1
a
“DJF” refers to winter months: December, January, February.
b
“MAM” refers to spring months: March, April, May; “SON” refers to autumn months: September, October, November.
c
“JJA” refers to summer months: June, July, August.
Curonian Lagoon drainage basin modelling and assessment of climate change impact 99
Table 7 Simulated inter-seasonal Nemunas and Minija river average (av), minimal (min), and maximal (max) discharge change for
P O
the pessimistic ( ) and optimistic ( ) scenario.
River River discharge change [ %]
Winter Spring Summer Autumn
av min max av min max av min max av min max
P
Minija 22 21 23 7 28 2 18 64 44 10 10 23
P
Nemunas 17 20 1 10 24 19 8 16 14 9 0.5 12
O
Minija 18 16 17 5 25 8 10 65 25 5 18 20
O
Nemunas 10 8 1 9 32 20 2 20 2 3 3 2
humidity will result in a stronger inter-seasonal fluctuation of run-off will increase by approximately 20% (see Table 7,
run-off. During the winter months, it is expected that the Fig. 5), although the SWAT overestimated minimal run-off
Nemunas River run-off will increase by 17% in the short term. for the Nemunas River in some cases, so this percentage could
The probable reason for this is the increased winter tem- be less.
peratures, which will result in earlier snowmelt. An increase For the Minija River, the effect of early snow melting is
in precipitation is also having a strong effect on run-off during more prominent; the increase in discharge will be approxi-
the winter months. The peak run-off for the winter will mately 22%. Minimum and maximum discharges for the win-
experience no significant change, whereas the minimal ter months will also increase by 21—23%. The strongest
Figure 5 Monthly simulated discharges of base, optimistic, and pessimistic scenarios for Minija and Nemunas rivers.
100 N. Čerkasova et al.
increase of discharges for both rivers is observed in February, for Minija. The strongest increase is observed during February
where the peak flow values are simulated to increase by more for the Nemunas River.
than 60%. During the spring months, a reduction in river flow is
During spring months, a 10% decrease in run-off is expected: 5% and 9% for Minija and Nemunas, respectively
expected for the Nemunas River and a 7% decrease is (Table 7). The maximal discharge will decrease by 8% and
expected for the Minija River (Table 7, Fig. 5). This is caused 20%, respectively, and the rivers' minimal discharges will
by the ice-melting season moving to the winter months. The decrease by even more: 25% and 32%. The strongest decrease
maximum spring discharges in Lithuanian rivers generally in discharge is observed in April for Nemunas, and in May for
take place during March to April, but in the light of current Minija (Fig. 5).
climate change, these events will happen earlier. The stron- During the summer months, a small increase in precipita-
gest decrease in discharge during spring months is observed in tion and a decrease in humidity were used in the optimistic
May for Minija, whereas for Nemunas, it is observed in April. scenario, with no change in the temperature. However, a
The summer months are expected to be warmer; this has a decrease in the rivers' run-off was simulated: 10% for Minija
significant negative impact on the discharge of rivers during and 2% for Nemunas (Fig. 5). This may be caused by the
this period. For Nemunas, it results in an 8% decrease in run- increased ET (evapotranspiration) and a higher water uptake
off, whereas for Minija, it results in an 18% decrease. The by plants. The highest reduction in discharge is observed in
minimal and maximal discharge values vary: for Nemunas, the period of June to July for both rivers.
peak flows during summer months will increase by 14%, but The autumn months will experience a small increase in
the minimal flow will decrease by 16%. For Minija, peak flows river discharge: 5% for Minija and 3% for Nemunas. The peak
will increase by 44% and minimal flows will decrease by 64%. and minimal flows will experience small fluctuations in both
The highest discharge change is expected during June to July cases. A general decrease in run-off during September, com-
for Nemunas, and in July for Minija (see Fig. 5). Even if in pared to the baseline scenario, is simulated for Minija, with a
some cases the model overestimated the values of peak flows gradual increase of flow in the following months. The Nemu-
and underestimated the minimal flows, the inter-seasonal nas River's discharge displays a steady increase during Sep-
differences between climate change and baseline scenarios tember and October, with the highest values occurring in
are significant. November. With this optimistic scenario, the annual dis-
Discharge changes in the autumn months are less affected charge in the near-term for the Nemunas River will increase
by climate change. During this period, the average discharge by around 5%, and by 2—3% for the Minija River.
will increase by 10% for Minija and 9% for Nemunas. Maximal
run-off will increase by 23% for Minija and by 12% for Nemu-
4. Discussion
nas. Minimal run-off will decrease by 10% for Minija and will
not change significantly for Nemunas (Table 7). In the autumn
The SWAT is a very useful tool for investigating climate
months, the average discharge will increase for Nemunas
change's effects on the drainage basin, assessing manage-
during the whole season, with no distinct patterns. The
ment strategies on a catchment area's hydrological and
Minija River's discharge, however, displays a decrease in
water-quality response and other different scientific and
the run-off during September, and a steady increase in the
practical uses. However, calibration and validation of the
following months, reaching the highest increase during
model is a key factor in reducing uncertainty and increasing
November (Fig. 5).
confidence in its predicative abilities, thus making applica-
The annual discharge in the short term for the Nemunas
tion of the model effective.
River will increase by around 7%, and Minija's will increase
The Curonian Lagoon drainage basin model was success-
slightly, by around 2—3%. These results confirm those of
fully calibrated and validated, although some improvements
similar studies (Kriaučiunienė et al., 2008; Meilutytė-Baraus-
to the results could be achieved in the future. During cali-
kienė and Kovalenkovienė, 2007; Rogozova, 2006), which
bration, the model simulations generally underestimated
indicate a slight increase in annual river run-off in the
high seasonal amounts of run-off for Minija, especially during
near-term and a change in flood behaviour during the spring.
the spring flood months of March to April. For both rivers, the
model underestimated the amount of discharge during the
3.2. The optimistic scenario's results months of June to August. This could be caused by some fitted
parameters of groundwater or base-flow; an overestimation
For the optimistic scenario, low values of precipitation and of plants' water uptake could also be the reason for these
temperature change were used to assess their effects on river occurrences. Further improvements to the model could
run-off. As expected, results of the optimistic scenario show assess the influence of each parameter on the run-off sepa-
smaller changes than those of the pessimistic one, although rately, for acquiring a better understanding of the river-
their tendencies remain the same (see Table 7). discharge governing processes for each sub-basin.
For both Nemunas and Minija, the expected river dis- During the climate scenario evaluation, both optimistic
charge will change mostly in the winter season: 18% for and pessimistic scenario simulations produced similar gen-
Minija and 10% for Nemunas. Minimal and maximal discharges eral patterns in changes to river discharge: a strong increase
during this season will increase by 16% and 17%, respectively, in the winter months, especially in February, a decrease
for the Minija River, while for Nemunas an increase of 8% in during spring and summer and a slight increase during
minimal discharge is expected, where the peak flows will autumn. It is noteworthy that even in the optimistic scenario,
remain at almost the same level. The increase in discharge is river discharges show a relatively high reduction during the
simulated during the entire season, with no distinct patterns spring months, meaning that the temperature threshold for
Curonian Lagoon drainage basin modelling and assessment of climate change impact 101
snowmelt can be reached even with a small increase in fill in the time-based gaps of monthly monitoring of the
temperature (see Fig. 5). Nemunas River's outlet, giving an idea of what kind of varia-
Different climate-change factors have influenced the tions occurred in the period of a month between two mon-
simulated changes in different ways. The relative humidity itoring expeditions. With further research and additional
change has an impact on river discharges through an increase calibration, this model can be used to simulate sediment,
in water uptake by plants and ET. The share of the forested pesticide and nutrient transport in the basin. The model
area in the Minija River's basin is approximately 21% (Kon- developed in this study can be linked to ecological-, biogeo-
tautas and Matiukas, 2010), and about 35% for the Nemunas chemical- and sediment-transport models for the Curonian
River's basin (Rimkus et al., 2013). Relative humidity can Lagoon. It can also support water-quality management stu-
affect the flow of water through the plant: the higher the dies of the Curonian Lagoon as well as scientific projects such
relative humidity, the more slowly transpiration occurs and as the ecological response of the Curonian Lagoon to differ-
vice versa. In the Curonian Lagoon drainage basin model, a ent load conditions through the Nemunas River, or detailed
reduction in relative humidity led to a reduction in river run- studies of carbon, nitrogen and phosphorus budgets.
off during the summer months. This was the case even in the
optimistic scenario, where higher precipitation values were Acknowledgements
used and the temperature values were not altered for this
period. Relatively high absolute changes in minimal and
The authors wish to acknowledge the support of the Research
maximal flows were simulated for the Minija River during
Council of Lithuania for the project “Promotion of Student
the summer months, especially in the pessimistic scenario.
Scientific Activities” (VP1-3.1-ŠMM-01-V-02-003). The authors
Minija is a river dominated by rain floods in the run-off
would also like to thank Klaipeda University's Open Access
balance. This factor becomes even more distinct in the light
Centre for Marine Research for providing data and computa-
of climate change, where heavy rain results in high local
tional resources.
increases of generated discharge. Approximately half of the
total run-off comes from rainwater; snow and groundwater References
comprise 22% and 25%, respectively, of the run-off. As a result
of the earlier melting of snow, these values are projected to
Abbaspour, K.C., 2011. SWAT-CUP2: SWAT Calibration and Uncertain-
change accordingly.
ty Programs Manual Version 2. User Manual. Department of
A general tendency for potential hydrological droughts
Systems Analysis, Integrated Assessment and Modelling (SIAM),
during the summer season is observed in both rivers. The
Eawag. Swiss Federal Institute of Aquatic Science and Technology,
Nemunas and Minija Rivers' basins lie under humid temperate Duebendorf, Switzerland.
climate conditions and cannot experience such water Arnold, J., Allen, P. , Bernhardt, G., 1993. A comprehensive surface
fl — —
shortages as appear in the tropical and mid-latitudinal arid groundwater ow model. J. Hydrol. 142 (1 4), 47 69.
Arnold, J., Kiniry, J., Srinivasan, S., Williams, J., Haney, E., Neitsch,
regions. Therefore, simulated dry periods and periods of low
S., 2010. Soil and Water Assessment Tool Input/Output File
streamflow are interpreted here as droughts, because of the
Documentation, Version 2009. Texas Water Resourc. Inst. Tech.
impact on wildlife and socio-economic factors (Rimkus et al.,
Rep. No. 365, 643 pp.
2013) — a decrease in crop yields, a reduction in overall
Arnold, J., Moriasi, D., Gassman, P. , Abbaspour, K., White, M.,
agricultural productivity, a massive increase in wildfires, an
Srinivasan, R., Santhi, C., Harmel, R., van Griensven, A., Liew,
intensification of tree defoliation, etc.
M.V., Kannan, N., Jha, M., 2012. SWAT: Model use, calibration and
The Minija River's sub-basin response to different climate-
validation. Am. Soc. Agric. Biol. Eng. 55 (4), 1419—1508.
fi
change factors was more signi cant than that of the Nemunas Balascio, C., Palmeri, D., Gao, H., 1998. Use of a genetic algorithm
River, which displayed some robustness to projected and multi-objective programming for calibration of a hydrologic
—
changes. As smaller rivers in the Curonian Lagoon basin's model. Am. Soc. Agric. Biol. Eng. 41 (3), 615 619.
CGIAR — Consortium for Spatial Information. SRTM Data Search
western boundaries display a high correlation (0.76—0.97) in
service, http://srtm.csi.cgiar.org/SELECTION/inputCoord.asp
run-off and synchronicity in flood seasons (Meilutytė-Baraus-
(accessed: February 2014).
kienė and Kovalenkovienė, 2007), it is assumed that the
Chaubey, I., Cotter, A., Costello, T. , Soerens, T. , 2005. Effect of DEM
changes in other, smaller rivers' sub-basins, such as the Deima
data resolution on SWAT output uncertainty. Hydrol. Process. 19
River's sub-basin, will be similar to those simulated for the
(3), 621—628.
Minija River's sub-basin.
Dubra, J., Červinskas, E., 1968. Freshwater balance of the Curonian
Possible future research directions may include improve-
Lagoon. Scientific Works of the High Schools. Geogr. Geol. 5, 19—
ment of the model's performance, a more thorough calibra- 26, (in Lithuanian).
tion and more detailed sensitivity and uncertainty analysis. In ELLE, PAIC, 2012. Development of methodics and modelling system of
addition, the completed Curonian Lagoon drainage-basin nitrogen and phosphorus load calculation for surface waters of
Lithuania. The 1st interim report of project, UAB Estonian,
model could be used to assess different climate-change,
Latvian & Lithuanian Environment and SIA Procesu analīzes un
water-management and agricultural-management scenarios.
izpētes centrs.
The SWAT can be coupled with other models that require a
Gailiušis, B., Kovalenkovienė, M., Jurgelėnaitė, A., 1992. Water
hydrological input, in order to assess different management
balance of the Curonian Lagoon. Energetika 2, 67—73.
problems and scenarios.
Gassman, P. , Reyes, M., Green, C., Arnold, J., 2007. The Soil and
As the SWAT model contains biogeochemical sub-models
Water Assessment Tool: historical development, applications, and
for nutrient transformation in its terrestrial and aquatic
future research directions. Am. Soc. Agric. Biol. Eng. 50 (4),
components, as well as plant growth and agricultural man- 1211—1250.
agement operations, the model developed in this study can Ghaffari, G., 2011. The impact of DEM resolution on runoff and
—
be upgraded to a full-featured drainage-basin model that can sediment modelling results. Res. J. Environ. Sci. 5 (8), 691 702.
102 N. Čerkasova et al.
Jakimavičius, D., 2012. Changes of water balance elements of the Moriasi, D., Arnold, J., Liew, M.V., Bingner, R., Harmel, R., Veith, T. ,
Curonian lagoon and their forecast due to anthropogenic and 2007. Model evaluation guidelines for systematic quantification of
natural factors (Ph.D. thesis). Kaunas Univ. Technol. accuracy in watershed simulations. Am. Soc. Agric. Biol. Eng. 50
Jakimavičius, D., Kovalenkovienė, M., 2010. Long-term water bal- (3), 885—900.
ance of the Curonian Lagoon in the context of anthropogenic National Centers for Environmental Prediction. Climate Forecast
factors and climate change. Baltica 23 (1), 33—46. System Reanalysis — Global weather database, http://
Kirtman, B., Power, S., Adedoyin, J., Boer, G., Bojariu, R., Camilloni, globalweather.tamu.edu/ (accessed: November 2014).
I., Doblas-Reyes, F. , Fiore, A., Kimoto, M., Meehl, G., Prather, M., Neitsch, S., Arnold, J., Kiniry, J., Williams, J., 2011. Soil and Water
Sarr, A., Schar, C., Sutton, R., van Oldenborgh, G., Vecchi, G., Assessment Tool. Theoretical Documentation. Version 2009. Texas
Wang, H., 2013. Near-term climate change: projections and Water Resourc. Inst. Theoretical Doc. 406, 618 pp.
predictability. In: Climate Change 2013: The Physical Science Rimkus, E., Stonevčius, E., Korneev, V. , Kažys, J., Valiuškevičius, G.,
Basis. Contribution of Working Group I to the Fifth Assessment Pakhomau, A., 2013. Dynamics of meteorological and hydrologi-
Report of the Intergovernmental Panel on Climate Change. Cam- cal droughts in the Neman river basin. Environ. Res. Lett. 8 (4),
bridge University. 045014.
Kontautas, A., Matiukas, K., 2010. Environmental Problems and Rogozova, S., 2006. Climate change impacts on hydrological regime
Challenges of the Minija River Towards a Sustainable Develop- in Latvian basins. In: Proceedings of the European Conference
ment of the River Basin. Watersketch Project Case Study Report. on Impacts of Climate Change on Renewable Energy Sources,
Coastal Research and Planning Institute, Klaipeda University. 137—140.
Krause, P. , Boyle, D.P., Base, F. , 2005. Comparison of different Rouholahnejad, E., Abbaspour, K., Vejdani, M., Srinivasan, R., Schulin,
efficiency criteria for hydrological model assessment. Adv. R., Lehmann, A., 2012. A parallelization framework for calibration
Geosci. 5, 89—97. of hydrological models. Environ. Model. Softw. 31, 28—36.
Kriaučiunienė, J., Meilutytė-Barauskienė, D., Rimkus, E., Kažys, J., United Nations Economic Commission for Europe, 2011. Second
Vincevičius, A., 2008. Climate change impact on hydrological assessment of transboundary rivers, lakes and groundwaters:
processes in Lithuanian Nemunas river basin. Baltica 21 (1—2), drainage basin of the Baltic Sea. Assessment report.
51—61. United Nations University. WaterBase project — Data for SWAT
Lin, S., Jing, C., Chaplot, V. , Yu, X., Zhang, Z., Moore, N., Wu, J., database, http://www.waterbase.org/download_data.html
2010. Effect of DEM resolution on SWAT outputs of runoff, sedi- (accessed: November 2014).
ment and nutrients. Hydrol. Earth Syst. Sci. 7, 4411—4435. Žilinskas, G., Jarmalavičius, D., Pupienis, D., Gulbinas, Z., Korotkich,
Meilutytė-Barauskienė, D., Kovalenkovienė, M., 2007. Change of P. , Palčiauskaitė, R., Pileckas, M., Raščius, G., 2012. Curonian
spring flood parameters in Lithuanian rivers. Energetika 53 (2), Lagoon coastal management study. Tech. Rep. Nature Heritage
26—33. Fund, (in Lithuanian).